tools//Claude Code//.claude//skills

Skills are reusable instruction packages invoked with `/name` in a Claude Code session: each skill is a directory inside `.claude/skills/` containing a `SKILL.md` file, with YAML frontmatter on top and markdown instructions below. The body is a prompt, and it can reference `$ARGUMENTS`, and `$0`, `$1`, `$2` for positional arguments.

Skills are reusable instruction packages invoked with /name in a Claude Code session: each skill is a directory inside .claude/skills/ containing a SKILL.md file, with YAML frontmatter on top and markdown instructions below. The body is a prompt, and it can reference $ARGUMENTS, and $0, $1, $2 for positional arguments.

Skills are not scripts: they tell Claude what to do with the existing tools, with no deterministic execution. The frontmatter takes these fields:

NameThe slash command name (commit gives /commit).

DescriptionWhat it does. Claude also uses it to decide whether to load the skill on its own when relevant.

Disable-model-invocationIf true, only the user can trigger it, never Claude on its own.

Argument-hintShown to the user as a placeholder (e.g. "[filename]").

Allowed-toolsRestricts which tools Claude can use while executing the skill.

The legacy path .claude/commands/<name>.md also works but lacks the frontmatter features.